abs->fabs fixes and other improvements

Author: Ithanil

examples/ex3/main.cpp

@@ -102,7 +102,7 @@ int main(){
 
     cout << endl << " - - - WAVE FUNCTION OPTIMIZATION - - - " << endl << endl;
 
-    const long NMC = 2000l; // MC samplings to use for computing the energy
+    const long NMC = 20000l; // MC samplings to use for computing the energy
     double * energy = new double[4]; // energy
     double * d_energy = new double[4]; // energy error bar
     double * vp = new double[psi->getNVP()];

src/ConjugateGradientTargetFunction.hpp

@@ -86,7 +86,7 @@ class ConjugateGradientTargetFunction: public NoisyFunctionWithGradient
         // compute direction (or gradient) to follow
         for (int i=0; i<_wf->getNVP(); ++i){
             grad_E[i] = 2.*( HOi[i] - H[0]*Oi[i] );
-            dgrad_E[i] = 2.*( dHOi[i] + abs(H[0]*Oi[i])*(dH[0]/H[0]+dOi[i]/Oi[i]) );
+            dgrad_E[i] = 2.*( dHOi[i] + fabs(H[0]*Oi[i])*(dH[0]/H[0]+dOi[i]/Oi[i]) );
         }
 
         if (_lambda_reg > 0.) { // compute the regularization derivative

src/NoisyStochasticReconfigurationOptimization.hpp

@@ -23,7 +23,7 @@ class NoisyStochasticReconfigurationOptimization: public WFOptimization{
     // optimization
     void optimizeWF(){
         // create targetfunction
-        NoisyStochasticReconfigurationTargetFunction * targetf = new NoisyStochasticReconfigurationTargetFunction(_wf, _H, getMCI(), _Nmc);
+        NoisyStochasticReconfigurationTargetFunction * targetf = new NoisyStochasticReconfigurationTargetFunction(_wf, _H, getMCI(), _Nmc, 0., true);
         // declare the Dynamic Descent object
         DynamicDescent * ddesc = new DynamicDescent(targetf);
         // allocate an array that will contain the wave function variational parameters

src/StochasticReconfigurationOptimization.cpp

@@ -40,7 +40,7 @@ namespace sropt_details {
         }
 
         // perform the integral and store the values
-        MPIVMC::Integrate(w.mci, w.Nmc, obs, dobs, true, true);
+        MPIVMC::Integrate(w.mci, w.Nmc, obs, dobs, true, false);
 
         // clear
         w.mci->clearObservables();
@@ -81,7 +81,7 @@ namespace sropt_details {
                 for (int j=0; j<nvp; ++j){
                     gsl_matrix_set(sij, i, j, OiOj[i*nvp + j] - Oi[i] * Oi[j]);
                     if (flag_dgrad) gsl_matrix_set(rdsij, i, j,
-                                   (dOiOj[i*nvp + j] + abs(Oi[i]*Oi[j])*( (dOi[i]/Oi[i]) + (dOi[j]/Oi[j]) ))
+                                   (dOiOj[i*nvp + j] + fabs(Oi[i]*Oi[j])*( (dOi[i]/Oi[i]) + (dOi[j]/Oi[j]) ))
                                    /  gsl_matrix_get(sij, i, j) );
                 }
             }
@@ -90,7 +90,7 @@ namespace sropt_details {
             for (int i=0; i<nvp; ++i){
                 gsl_vector_set(fi, i, H[0]*Oi[i] - HOi[i]);
                 if (flag_dgrad) gsl_vector_set(rdfi, i,
-                               (abs(H[0]*Oi[i])*( (dH[0]/H[0]) + (dOi[i]/Oi[i]) ) + dHOi[i])
+                               (fabs(H[0]*Oi[i])*( (dH[0]/H[0]) + (dOi[i]/Oi[i]) ) + dHOi[i])
                                / gsl_vector_get(fi, i)  );
             }
             // invert matrix using SVD
@@ -127,7 +127,7 @@ namespace sropt_details {
                 for (int k=0; k<nvp; ++k){
                     foo = gsl_vector_get(fi, k)*gsl_matrix_get(Isij, k, i);
                     grad_E[i] -= foo;
-                    if (flag_dgrad) dgrad_E[i] += abs(foo) * ( gsl_vector_get(rdfi, k) + gsl_matrix_get(rdsij, k, i) );  // not correct, just a rough estimation
+                    if (flag_dgrad) dgrad_E[i] += fabs(foo) * ( gsl_vector_get(rdfi, k) + gsl_matrix_get(rdsij, k, i) );  // not correct, just a rough estimation
                 }
             }